Trace Element Studies of the Arkansas Novaculite Please do not cite with out permission. Un-redacted version may be available upon request Trace Element Studies of the Arkansas Novaculite Thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts By Kristin D. Scarr, B.S. Mercyhurst College, 2004 May 2008 The University of Arkansas ABSTRACT Novaculite color and texture can be widely variable. This pilot study assessed how much variability in trace element concentration is detectable throughout the formation as well as within individual sample areas and compared two analytical procedures for their suitability for source characterization in archaeology. Trace-element analysis was conducted through the use of instrumental neutron activation analysis (INNA) at the University of Missouri Research Reactor and scanning electron microscopy-energy dispersive x-ray spectrometry (SEM-EDS) at the University of Arkansas ESEM laboratory. The raw data from INAA were subjected to principle component analysis with encouraging results. Two distinct populations, ArkansasOklahoma and Texas, have emerged in the analysis. Three smaller groups that correspond with the sample areas can also be seen in the Arkansas-Oklahoma population. While the results are preliminary they do provide a glimpse of the potential variation present throughout the formation at large. Based on these results further study is recommended to further characterize the formation as a whole, including the novaculite outcrops in Oklahoma and Texas. A comparison of the results from the INAA and EDS methodologies was also undertaken to determine whether EDS was a valid method for trace element characterization. The EDS results from this study are not amenable to the same statistical analyses that the INNA were. Therefore, further testing using EDS is recommended to fully examine the utility of this methodology. This thesis is approved for Recommendation to the Graduate Council Thesis Director: _____________________________________ Marvin Kay Thesis Committee: _____________________________________ Mary Beth D. Trubitt _____________________________________ Walter L. Manger _____________________________________ Kirstin T. Erickson ACKNOWLEDGEMENTS First and foremost I would like to thank Dr. Michael Glascock, Matthew T. Boulanger and the staff at MURR for granting partial funding to Mary Beth Trubitt for INAA, and for providing me with a comprehensive report of the data. Their work is the basis for most of the conclusions made in this study. They were extremely helpful throughout the process providing me with the necessary data and source material and were always willing to answer questions. I thank Dr. Mary Beth Trubitt for helping me to develop my thesis topic. I would also like to thank her for helping me collect samples and for providing reference material. Dr. Trubitt and the Arkansas Archaeological Survey will use the data recovered during the analysis to determine if a larger more intensive survey should be conducted. I would like to thank the Arkansas Archaeological Survey, Tom Green for funding the remaining portion of the INAA and for providing me with the funds to collect my samples. I also thank the University of Arkansas’ department of Anthropology for providing funds for EDS analysis. I thank Greg Butts, Director of Arkansas State Parks and Bill Saunders the superintendent of for granting me a permit to collect samples in the park. I would also like to extend thanks to George Sabo, Jerry and Leslie Walker for their help with grammar and spelling questions, Lelia Donat and Marion Kunetka in the AAS registrars office for their help with quarry site information, and Aaron Lingelbach for helping me to break up rocks outside in the cold. I would like to extend my gratitude to Meeks Etchieson, Heritage Program Manager for the Ouachita National Forest for his assistance in sample collection at and for his donation of samples from Oklahoma. I would also like to thank Dr. Walter Manger for all of his help in developing ideas for my thesis topic and v for providing me with geological tools and literature. I want to thank Alan Toland for conducting the EDS testing of my samples. I also want to thank Dr. Marvin Kay for his thorough editing and advice for improving my writing. I would like to thank Charles Frederick for his assistance in figuring out the Texas novaculite issue. Finally, I want to thank Laura K. James for her support and her willingness to edit at all hours and my family for their encouragement. vi TABLE OF CONTENTS ABSTRACT ACKNOWLEDGEMENTS INTRODUCTION ii v 1 CHAPTER 1: Geological Background Color and heat treatment 7 15 CHAPTER 2: History of Research Lithic raw material characterization Novaculite characterization studies 20 26 29 CHAPTER 3: Methodology Testing Methodology 31 41 CHAPTER 4: Data and Results EDS (energy dispersive x-ray spectrometry) INAA (instrumental neutron activation analysis) 44 45 48 CHAPTER 5: Concluding Remarks 59 References Cited 64 Appendix A: Sample collection database Appendix B: EDS raw data Appendix C: Neutron Activation Analysis Report from MURR, Matthew T. Boulanger and Michael D. Glascock 70 77 vii 85 FIGURES Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Map of the Arkansas Novaculite formation. Map of the Ouachita orogenic belt. Stratigraphic column of the Arkansas Novaculite Formation Keller et al. (1985) proposed metamorphic zones. Miser (1943) Proposed metamorphic zones. Indian Mountain quarry debris photographs. Caddo Gap road cut photographs. Sample location map Arkansas and Oklahoma. EDS graphical data INAA PCA graph. INAA PCA graph. Thorium-Uranium graph Europium-Uranium graph. Chromium-Manganese graph. Hierarchical cluster analysis; Ar-Ok and Tx. Hierarchical cluster analysis; Ar and Ok. 9 10 11 17 18 23 36 40 45 52 53 54 55 56 57 58 TABLES Table 1 Table 2 List of novaculite origin theories from various authors, after Sholes 1978. Sample list with provenance. viii 14 34 INTRODUCTION 1 “Ascending a very lofty hill composed entirely of [novaculite], we found several large pits, resembling inverted cones, some of which were from 20 to 30 feet deep and as many in diameter, the insides and bottoms of which were covered with chips of this beautiful mineral, some white, some carmine, some blue, and many quite opalescent. […] These were undoubtedly the quarries from whence the Indians, when they possessed the country, obtained the materials for making their arrow heads and spears, […]” (Featherstonhaugh 1844:111). Featherstonhaugh (1844) was the first to describe the Arkansas Novaculite during his travels through the southern states. This vivid description set the tone for all subsequent studies of the peculiar material. The purpose of this study is to examine the elemental composition of the Arkansas novaculite in order to determine the nature of intra-formational variation. It will also establish whether energy dispersive x-ray spectrometry (EDS) is a suitable alternative to instrumental neutron activation analysis (INAA) for the chemical characterization of novaculite. INAA is a destructive technique rendering samples unusable, while EDS allows samples to be returned. The Arkansas novaculite formation is composed of several different layers of rock, inter-bedded chert, shale, novaculite, chert-clast conglomerate and sandstone; identified as five geologic members from bottom to top: lower chert and shale, lower novaculite, middle chert and shale, upper novaculite and upper chert and shale (Sholes 1978). These members are not present in all outcrops and their thickness varies throughout. For the purpose of this study I use Mark A. Sholes’ definition of novaculite. He defines it as “ a siliceous rock composed of polyhedral grains of microquartz which contains almost no chalcedony and which can usually be distinguished from chert by its gritty rather than smooth fracture 2 surface” (Sholes 1978:v). The study focuses on novaculite outcrops and aboriginal quarries in the Ouachita Mountains of central Arkansas. A control sample from the western extent of the formation in eastern Oklahoma was also included. The study of novaculite by geologists mostly concerns its origins and often includes trace element analysis (Cornish 1997; Doerr 2004; Griswold 1892). While archaeological studies focus on novaculite quarry sites and artifacts, a few have included novaculite for comparison with other siliceous materials (Ives 1995; Luedtke 1992). One study (Flenniken and Garrison 1975) investigates the effects of thermal alteration on the microcrystalline structure of novaculite. One goal of this study is to compare destructive and non-destructive analytical methods. When studying artifacts, minimizing the destruction of the samples is essential. Neutron activation (NAA) and its modern equivalent INAA, has long been the chosen technique for trace element identification. Its long history of use and ability to detect small concentrations of elements make it the preferred method for all manner of archaeological analyses. Unfortunately, this method destroys test samples through crushing and irradiation. Non-destructive techniques for characterizing lithic raw materials would preferable. Energy dispersive x-ray spectrometry (EDS) may be one such methodology. EDS is generally non-destructive in that small specimens can remain intact and sample preparation is often minimal. If the results of EDS were comparable those from INAA, we could forgo the destruction inherent in neutron activation analyses. The Arkansas Novaculite is a unique raw material that has been utilized by both prehistoric and historic peoples. While similar to chert in its general chemical 3 composition, namely SiO2, novaculite is often considered to be a separate entity. Novaculite tends to lack chalcedony, which gives it a more granular texture. It also tends to be more translucent than chert, especially around the edges of the specimen. However, it is often hard for someone unfamiliar with novaculite to recognize it as separate from other siliceous rocks. Difficulty in classification has lead researchers to seek out alternatives to visual identification of novaculite. If INAA and EDS are able to identify a significant amount of variation between and perhaps within each sample location, chemical testing can be extended to artifactual materials. Through the testing of and comparison to cultural material, we will be able to create a database of known quarry locations and associated artifacts. The database will also be extremely useful for the reconstruction of past life ways, especially those involving craft production, trade and exchange. The text is divided into five chapters: Geologic Background, History of Research, Methodology, Data and Results and Concluding Remarks. The fifth chapter contains interpretation of the results while answering several important research questions. 1) Can we detect intra-formational variation in novaculite? If variation exists, is it significant? While results are preliminary, a significant amount of variation was detected within the Arkansas Novaculite formation in both the INAA and EDS results. 2) Is the variation enough to differentiate between source areas, especially those in close proximity? Statistical analysis was able to identify several groupings within the data that correspond with the sample locations. However, the variation within each sample 4 area is greater than that within the entire formation. This disparity is likely due to the small number of samples. 3) Can we use these data to identify the particular quarries as raw materials sources for artifact production? Will we be able to determine the source of individual artifacts? The results of this study are insufficient to determine individual source locations. Further testing of a larger more representative number of samples is required before any conclusions can be drawn. 4) Is there a difference between the results from INAA and EDS? If so what is the difference and what are the implications for future research? EDS identifies a different set of elements than INAA. They have many of the same elements in common but there are a few important differences. This has a significant impact on the results in that elements that are not shared by both techniques cannot be included in statistical analyses. The incompatible elements could be the key to determining the difference between outcrops or sources but would be ignored by the analysis. Another issue with EDS testing was the lack of proper recording during testing. If this method is to be used in the future, tighter control of sample testing, including the recording of time intervals and observations, by the technician is recommended. It would also be important to establish parameters that will provide reliable and comparable results. The current EDS results are not directly comparable to those from INAA. That being said the two techniques have produced results that are encouraging for future research into this topic. However, INAA is a more reliable well-tested methodology for trace element characterization. 5 EDS may still be a useful technique, but further and more intensive testing must occur before it can be utilized as an alternative to INAA. 6 CHAPTER 1: GEOLOGICAL BACKGROUND 7 The Arkansas Novaculite is located in the Ouachita Mountains and stretches from central Arkansas to eastern Oklahoma (figure1). Novaculite outcrops can also be found in west Texas within the Marathon uplift area [figure 2](Sholes 1978). These Texas outcrops are part of the Caballos Novaculite, rather than the Arkansas. The novaculite formations and associated outcrops lie along the Ouachita structural belt, the result of orogeny that began in the Paleozoic and continued into the Mississippian (Aber 2003). A map by Grunig (1974) [figure 2] shows the orogenic belt and the corresponding areas of uplift. The Arkansas Novaculite was deposited during the Upper Devonian and Lower Mississippian and rest conformably above the Missouri Mountain Formation in most of the outcrop locations (McFarland 2004: 21). The formation is composed of five members. Sholes (1978) identifies them form bottom to top: lower chert and shale, lower novaculite, middle chert and shale, upper novaculite and upper chert and shale (see figure 3). The members are not present in all outcrops and their thickness varies. The formation thickness can extend from 900 feet in the south and to as little as 60 feet in the north (McFarland 2004:21). The Arkansas Novaculite was first described by D.D. Owen in 1860. Novaculite has been redefined based on individual outcrops rather than the formation as a whole. As a result, the definition of novaculite is continuously debated in the literature. Miser and Purdue (1929:49) defined novaculite as “gritty, fine-grained, homogenous, highly siliceous rock, possessing a conchoidal or sub-conchoidal fracture and being translucent in thin edges.” 8 9 10 11 Holbrook and Stone (1979) depict novaculite “as a homogeneous mostly white or lightcolored rock, translucent on thin edges with a dull to waxy luster comprised almost entirely of microcrystalline quartz.” For Lemley (1962) “ novaculites are of two classes, known as the Arkansas and the Ouachita. The former, a true novaculite, is a fine-grained, homogeneous stone of waxy luster, translucent, with a marked conchoidal fracture. The Ouachita stone is less dense, coarser, less translucent and lacks the waxy luster […] and has the dead appearance of unglazed porcelain”(1962). Keller et al. (1977:843) have “proposed that novaculite” should be used as the official term for rocks that are “a thermally metamorphosed siliceous rock exhibiting polygonal, triple point texture.” There are those who disagree with this definition because it cannot be extended to other parts of the formation. In his 1978 dissertation, Mark A. Sholes defines novaculite as “ a siliceous rock composed of polyhedral grains of microquartz which contains almost no chalcedony and which can usually be distinguished from chert by its gritty rather than smooth fracture surface”(v). Sholes (1978:61) points out that since the rocks at Caddo Gap do not exhibit metamorphism the Keller et al. (1977) definition, among others, “eliminates [them] from being called novaculite. The long use of novaculite for these rocks makes this definition unacceptable […] Because of the widespread and imprecise use of novaculite as a rock name, it is not considered useful to redefine novaculite on the basis of a metamorphic texture […]” Many interpretations of novaculite deposition and diagenesis center on biogenic siliceous sedimentation and the inorganic replacement of carbonates through groundwater flow as possible sources for the silica. Biogenic sediment contains the microscopic remains of the unicellular, silica-secreting plankton, radiolarians and shards of sponge 12 spicules. Explanations tend to center on an underwater origin regardless of where the silica originates. A second interpretation to biogenic sedimentation is the precipitation of volcanic sediments into the ocean (Honess 1923; Goldstein and Hendricks 1953; A.R. Niem 1977), and is starting to be widely accepted. Walter L. Manger (personal communication 2008), a geologist at the University of Arkansas agrees with the idea of a volcanic origin, citing the sheer volume of silica in the novaculite as proof of this. It has also been suggested that due to the nature and extent of the deposits, the variable texture of the novaculite, as well as the lack of fossil material and chalcedony it owes its current form to widespread metamorphic activity (Cornish 1997). The nature of the sediments before the metamorphic activity is debatable but the evidence of it is generally accepted. Sholes’ (1978) summary of the varied theories of novaculite origins is reproduced in Table 1. They are divided into five categories: 1) replacement of limestone by silica; 2) diagenetic or metamorphic alteration of clastic quartz; 3) inorganic replacement of silica; 4) organic precipitation as skeletal particles, subsequently recrystallized or altered; and 5) sea-floor alteration of volcanic ash. The Arkansas Novaculite also comprises layers of sandstone and shale. Sholes (1978) describes it as “ composed predominately of inter-bedded chert, shale, and novaculite, with minor chert-clast conglomerate and sandstone.” There is also a small amount of what are called “accessory minerals”, the most common being “ tiny grains of hematite and pyrite”(Sholes 1978). There are also trace amounts of calcium and manganese in much of the formation and larger deposits of manganese, in the form of “nodules, pockets and short, irregular veins […]” (Sholes 1978). Hydrothermal activity 13 in the Ouachitas may also have contributed to the influx of elements such as As, Hg, Se, Sb, and Ni found in the lower member of the Arkansas Novaculite (Cornish 1997:45). Concentrations of these minerals in the upper member of the formation are unknown. Table 1: Summary list of Arkansas Novaculite origin theories modified from Sholes (1978) after McBride and Thomson (1970:76). Owen (1860): sandstone altered by heated alkaline siliceous water. Branner (in Comstock 1888): novaculite is selectively metamorphosed bedded chert. Comstock(1888): alteration of quartz (sandstone implied) in place by hot water. Griswold (1892): very fine clastic quartz. Rutley (1894): siliceous replacement of dolomite or dolomitic limestone. Hinde (in Rutley 1894): probably organic silica. Derby (1898): replacement of limestone by silica. Weed (1902): chemical precipitate in deep sea. Van Hise (1904): organic precipitate, now recrystallized. Honess (1923): in part silicified and devitrified volcanic ash, but mostly chemical precipitate without the aid of organisms. Miser and Purdue (1929): chemically precipitated. Henbest (1936): at least in part organically precipitated Hendricks et al. (1937): initially siliceous deposit derived in part from siliceous organisms. Harlton (1953): replacement of limestone by ground water upon rupture of rocks during tectonism Goldstein and Hendricks (1953): submarine alteration of volcanic ash to produce opaline or ‘isotropic’ silica; minor contribution from siliceous organisms; diagenetic recrystallization to chalcedony and cryptocrystalline silica. Park (1961): recrystallized amorphous silica deposited by organisms. McBride and Thomson (1970): diagenetically altered organic silica in deep water. Folk (1973): novaculite formed by silicification of spiculitic siliceous, calcareous, or evaporitic sabkha or lagoon sediment and bedded chert by organic deposition of silica in brackish open marine water enriched in silica by rivers. Keller et al.(1977): novaculite formed by metamorphism of chert. Lowe (1976): novaculite formed from pelagic biogenic silica and chert and shale members formed in part, from turbidities. 14 Color and Heat Treatment Novaculite comes in a variety of colors. Shades of white and gray are the most common and can often be found on the same piece. Pink is another common color variety often found with white and gray. There are many other varieties, such as black or very dark gray, tan, light brown, red, green, yellow and orange. Most of these colors can be seen within the same rock. The colors can be marbled or mixed together and can be layered/adjacent to one another with gradual or abrupt color transitions. Jenney described the color variety in his 1891 article “ The rock is white, yellowish, or bluish white in color, breaking readily with a smooth conchoidal fracture” and continues to describe other quarry areas as having “chips of pink, red or white novaculite, rarely dark-colored or black and always having a fine-grained structure”(Jenney 1891: 316-317). Jenney’s color descriptions also provide important textural information. In most geological and archaeological descriptions of the Arkansas novaculite, texture and luster are described as waxy or fine-grained. As with any siliceous material, texture can vary widely, especially if it was formed in association with several different sedimentary environments. Novaculite is often described as having a grainer texture than most siliceous rocks. This is due to the molecular structure of the material. Novaculite is composed of microcrystalline quartz and contains almost no chalcedony. Griswold (1890:187), a geologist from the Geological Survey of Arkansas, cites “microscopic examination shows that the soluble silica of chert is in the form of chalcedony, while novaculite is entirely without silica in this form.” Chalcedony is quartz that forms hairlike structures instead of crystals. This fiber bundle structure creates a more even and vitreous or glassy texture. The length of time during which sediment is laid down can 15 also affect the characteristics of the resulting formation. The variables that affect texture and luster also have an effect on the fracture characteristics of the stone. The more homogenous or vitreous varieties will exhibit true conchoidal fracture and are more likely to be selected for tool production. Metamorphism can also alter the original texture of a material. Keller et al. (1977) utilized SEM micrographs to determine the texture of novaculite. They concluded that, “the Bigfork Chert and Arkansas Novaculite formations, where exposed to raised temperatures in the Ouachita Mountains of Arkansas, especially in the Hot Springs-Little Rock region, were recrystallized to acquire a polygonal, triple-point texture” (Keller et al. 1977:842,843). The metamorphism appears to be a localized event occurring in two main areas of the formation, according to Miser (1943), without encompassing it entirely. In figures 4 and 5, two versions of the metamorphic zone have been depicted on a map of the formation. The Caddo Gap novaculite sampled for this study lies within a proposed non-metamorphic zone. The remaining samples from and possibly those from Oklahoma, were collected from the areas of proposed metamorphism. 16 17 18 In contrast, Keller et al. (1977) argue that the rocks at Caddo Gap were altered by metamorphism. They extend this to the entire formation. Their conclusion is based on crystal size, which is known as the crystallinity index. However, Sholes (1978) suggests that the same results can be obtained for non-metamorphic rocks. He maintains that it can occur due to differences in the original sediment or from overlapping/stacked grains that alter the appearance of the lattice structure. Metamorphic alteration of rocks can also have an effect on the elemental composition of a material. Elements can be destroyed by heat and pressure, while others may be added from intrusive liquids or materials. This activity can have an impact on characterization studies, allowing researchers to differentiate novaculite specimens on a sub-regional basis. 19 CHAPTER 2: HISTORY OF RESEARCH 20 The literature dedicated to lithic raw material sourcing studies is vast. The same can also be said of novaculite, which has been widely discussed and documented by archaeologists and geologists. The following is a brief overview of some of the relevant literature. George W. Featherstonhaugh first mentioned the Arkansas novaculite in his manuscript Excursion through the Slave States in 1844. During his survey of the land between the Red and Missouri rivers, Featherstonhaugh visited a few of the novaculite quarries in the area and recorded his observations. In 1892, L. S. Griswold, one of the Arkansas staff geologists, devoted an entire report to novaculite as a whetstone material. He also mentioned prehistoric quarries and the “so-called Spanish diggings,” now known as the quarries in the area (Etchieson 1997). Geologists have been studying novaculite and its formation for decades, formally cementing the name in the literature. Griswold (1890) discusses the material in length and while he was not the first to describe novaculite, his text was the earliest extensive treatment of the subject. Geologists have continued to maintain their interest in the material throughout the years. Honess (1923), Miser (1943), Lemley (1962), Lowe 1974), Folk and McBride (1976&1977), Keller, Viele and Johnson (1977), Keller, Stone and Hoersch (1985), Sholes (1977,1978) among others, have all contributed to the understanding of the novaculite and its geologic properties. In the mid to late 1880’s, William Henry Holmes explored aboriginal quarries for his manuscript Handbook of Aboriginal American Antiquities. Holmes described several quarries in the Ouachitas including the ‘Great Workshop’ (3GA22). Holmes (1974 [1919]: 196) described the novaculite quarries in Arkansas as “possibly 21 even more extensive than those of Ohio.” He also discussed the character of the deposits and even remarked on the “ […] enormous accumulations of shop refuse [that] have begun to [descend] on the interior of the mine”(Holmes 1974 [1919]: 196). Quarry debris, including abandoned items in various stages of preparation, litters the mountaintop; a sight that is hard enough to capture in a photograph let alone in writing (figure 6). In an article published in 1891, W. P. Jenney recounts his revisit to some of Holmes’ quarries including 3GA48 near The outcrops and quarry debris extend for miles along the ridge, as Jenney states, “ They consist of a number of shallow excavations upon the broad, rounded crest of the divide, covering a belt three hundred to six hundred feet in width, […] As far as I followed the divide—for a distance of one and a half miles --- these workings continued, and are reported to extend, with breaks at intervals, an extreme distance of four miles southwesterly from this point” (Jenney 1891:316). In the following decades the Arkansas Novaculite and its quarries were largely ignored. Meeks Etchieson (1997:4) states that up until the 1970’s “ very little effort was expended in examining or investigating the novaculite quarries. At most they were merely cited […] as needing further study.” 22 In 1974, Charles Michael Baker devoted a portion of his master’s thesis research on the Arkansas novaculite quarries. After visiting several Oklahoma quarries studied by W. H. Holmes, Baker was inspired to locate similar sites in Arkansas. As the investigation progressed, he was struck by the lack of information regarding these sites. “ The reports made clear that Arkansas’ archaeological resources included some very significant prehistoric novaculite quarries. However, to my surprise, none of the professional archeologists at the University of Arkansas was very familiar with these sites […] Thus, I proposed the ‘rediscovery’ of the novaculite quarry sites for the purpose of elucidating the [nature] of the importance of novaculite as a raw material used by prehistoric inhabitants of southern Arkansas” (Baker 1974: 6). Baker then conducted an inventory of Hot Springs National Park. He relocated Holmes’ Great Quarry atop along with several other smaller pits. Quarry pits were located during surveys in the area of the Caddo Gap road cut, in the towns of Bismark and Malvern and in State Park. The quarries atop Mountain were also surveyed and a minor excavation was conducted by Baker to determine the nature of the extensive midden deposits. In 1975, J. Jeffery Flenniken and Ervan G. Garrison conducted experiments on novaculite to determine if evidence for heat-treatment could be easily detected. The results of their experiments provided researchers with a simple technique for the identification of thermal alteration in novaculite and paved the way for similar studies of other siliceous materials. In her 1992 book, An Archaeologist’s Guide to Chert and Flint, Luedtke used novaculite as one of her test materials. Novaculite and other chert types are mentioned throughout the text to illustrate the various concepts that she presents. 24 Luedtke examined her samples with scanning electron microscopy, neutron activation and atomic absorption. Novaculite was also one of the central topics of Erica Doerr’s (2004) master’s thesis. While the text is mainly geological, Doerr considered its aboriginal use and included two samples of novaculite from the Hot Springs Arkansas area. Doerr had her samples analyzed in the environmental scanning electron microscopy (ESEM) laboratory at the University of Arkansas using EDS. The results of her analysis provide more questions than answers especially concerning “the method of data collection and the need for better statistical approaches.” (Doerr 2004:66). She does not include any information regarding length of testing interval for each sample or what power the instrument was set to during the tests. Without this, it is not possible to compare the results with this study. The most recent studies concerning novaculite have been undertaken by Mary Beth Trubitt of the Arkansas Archaeological Survey. Trubitt has conducted several surveys, quarry hikes and excavations that directly involve novaculite (Trubitt 2005a, 2005b, & Trubitt et al. 2004). In 1996, Trubitt and several other archaeologists from the Arkansas Archaeological Survey entered into a “cost-share agreement [with the U.S. Forest Service] to develop a detailed research design for investigating the novaculite quarry sites on lands in the Ouachita National Forest” (Trubitt et al. 2004:17). My thesis is a direct outgrowth of this research design and the contributions of Dr. Mary Beth Trubitt. Lithic Raw Material Characterization Archaeologists and geologists have been utilizing elemental analysis for several decades. Every year the technology evolves and new procedures and instruments are 25 developed. Because of this work, there is a wealth of information available for those interested in lithic sourcing studies. There are countless studies involving lithic source analysis, I have included a few examples to put my study in perspective. In Speakman and Glascock’s (2007) article Acknowledging Fifty Years of Neutron Activation Analysis in Archaeology, we are given a short history of the technique and the impacts it has had on archaeology as a discipline. One of the earliest characterization studies utilizing NAA was conducted in 1954 by J.R. Oppenheimer, Richard Dodson and Edward Sayre on Old World pottery (Speakman and Glascock 2007). According to the authors, “NAA [has] emerged as one of the most powerful and widely applied analytical techniques for chemical characterization and provenance-based research of ceramics, obsidian, chert, flint, basalt, glass, metals and other archaeological and historical materials”(Speakman and Glascock 2007:180). One of the major players in the world of neutron activation (as well as other chemical characterization methods) is the University of Missouri Research Reactor or MURR. MURR has been involved in many characterization studies over the years and maintains several databases of the work they have conducted including obsidian, ceramic and chert. There are several other research reactors in the U.S. and Canada conducting work on raw material identification such as the Smithsonian-NIST partnership in Washington D.C. and the SLOWPOKE reactor at the University of Toronto that closed in 1998 (Blackman and Bishop 2007; Hancock et al. 2007). In 1978, Barbara Luedtke conducted trace element analysis on a variety of chert types in the midwest utilizing neutron activation analysis. Her study is mostly an exercise in the viability of trace element characterization and presents guidelines for conducting 26 this kind of research. Because of her pioneering efforts, Luedtke is one of the most cited authors in the world of chemical characterization. She has written several articles as well as a book on the subject. Ludetke’s (1992), An Archaeologists Guide to Chert and Flint is an excellent resource for those who wish to conduct raw material sourcing. The book contains the proper methods for conducting an accurate and comprehensive source analysis. A major issue in quarry sourcing analyses is that of proper sampling procedures. She stresses that “a coherent sampling scheme must be used, and it should be designed to include the full range of physical variation for the chert type”(Luedtke 1978:422) She goes on to say that “source samples should come directly from geological deposits” and that “it is better to determine the full range of variation for the source and then find where the artifact values fit along that range”(Luedtke 1978:422). She also provides essential geological data on siliceous rocks, such as the processes of chert genesis and the details of silica formation. Tim Church’s (1994) book Lithic Resource Studies: A Sourcebook for Archaeologists is a comprehensive treatment of raw material sourcing. Like Luedtke, Church includes the methods and procedures that can be applied to almost any lithic sourcing study. The majority of the text is devoted to a large, mostly annotated bibliography that contains a variety of references that deal with the subject at hand. Lithic raw material characterization has really begun to blossom in the past few decades. While it may seem that the field is over saturated, the truth is that there is still much work left to be done. Siliceous rocks are by nature highly variable. The questions of their origins are still hotly debated and there is much we still need to study and define. Color, texture and chemical signatures can vary greatly within individual quarries as well 27 as throughout the formation at large. It is this ‘land of confusion’ that has fostered characterization studies all over the world. It is the hope of archeologists, archaeometrists and their geological counterparts that one day we will have a comprehensive database of siliceous rocks of all types. Most lithic characterization studies can be grouped under three main headings: 1) Obsidian and other natural glasses, 2) Chert, which includes other similar material names such as flint, jasper, chalcedony, agate, rhyolite, quartzite, novaculite and other miscellaneous silicates, 3) groundstone materials such as sandstone, granite, limestone and any other stone that may have been used by prehistoric Native Americans. Most studies focus on the first two, as groundstone materials are often taken for granted. I have included a short discussion and several references of the first two categories to place my own study into perspective. A commonly studied raw material is obsidian from many sources and is found at archaeological sites all over North and South America. Identifying the source of the obsidian is essential to understanding the prehistoric socio-economic trade and exchange patterns. Many obsidian characterization studies have utilized neutron activation analysis. Ambroz and Skinner (2001) submitted their obsidian samples to MURR for INAA and even collaborated with Michael Glascock, the director of the facility, in their publication of the results. Energy dispersive x-ray spectrometry (EDS) has also been used to analyze obsidian. Acquafredda et al. (1999) utilized the technique in their characterization of Mediterranean obsidian because of its non-destructive nature. Countless characterization studies have been conducted to determine the origins of chert artifacts. These studies span several decades and include sites all over the world. 28 Barbara Luedtke (1978), one of the early supporters of sourcing studies, utilized NAA and several varieties of chert to illustrate important concepts essential to trace element analysis. She points out that color is not necessarily correlated with chemical variation, and stresses the importance of proper sampling. Hoard et al. (1993) used INAA to analyze geological and artifactual samples of chert and chalcedony from the White River Silicate group in the Great Plains. Julig (1994) also utilized INAA in his study of Great Lakes chert sources and artifacts at the University of Toronto’s SLOWPOKE reactor facility. Novaculite Characterization Studies Archaeological characterization studies of Arkansas novaculite began in 1984 with D. J. Ives. He included eight samples of novaculite in his comparison of midcontinental sources with Crescent Hills (Missouri) chert through neutron activation analysis. The novaculite was distinguished from the rest of the chert samples because of its low level of Na and Cr concentrations and a high amount of Zr and Ce (Ives 1984). In 1992, Barbara Luedtke included 26 samples of novaculite tested by neutron activation analysis for her comprehensive guide on the subject of raw material studies. She found chemical differences between the novaculite samples collected from different sample areas. The samples that she tested from Caddo Gap had higher values than the rest of the novaculite for all of the elements except for Fe, U, and Sb. Luedtke attributes this to hydrothermal and metamorphic activity which “apparently resulted in a rather thorough flushing of many elements from Arkansas novaculite, but enrichment of some metals”(1992:60). 29 The geological community has also chemically characterized novaculite. In 1890, Griswold included a chemical analysis in his treatment of the stone. He did not, however, mention the technique used for analysis. Holbrook and Stone (1979) also analyzed the chemical composition of the novaculite and determined that it was 99% SiO2. Erica Doerr (2004), a geology master’s student at the University of Arkansas, included two samples of novaculite in her treatment of the Mississippian chert sources. She analyzed her samples using (EDS) energy dispersive x-ray spectrometry (2004). She does not mention the novaculite specifically in her results or conclusion, but she does include the raw data and appears to have tested the samples three times each. She does not describe her testing methodology in depth and it is uncertain which sample is represented by the six data sets labeled as novaculite. C.S. Cornish (1997), a geology masters student from Austin State, analyzed samples of novaculite from the Magnet Cove area and the Caddo Gap road cut, using atomic absorption spectrometry. He utilized the results to determine the possible origin of the novaculite deposits. 30 CHAPTER 3: METHODOLOGY 31 Five sites were selected and sampled for this study. Four of the sample localities are found in the central Arkansas Ouachita Mountains in and around the Hot Springs area. Two of the four sample sites are located Mountain and have been assigned archaeological site designations, 3HS603 and 3HS69. The fifth site is located in eastern Oklahoma on a mountain to the southwest of the Big Hudson creek. A representative collection of novaculite was gathered from each of the sample sites (3HS603 and 3HS69 on [3GA48], and Caddo Gap) in an attempt to characterize the potential variation in each area. Two pieces of novaculite were collected from Big Hudson Creek Mountain in Oklahoma. Due to limited budget and time, only twenty samples were submitted for testing. The initial proposal design called for four samples to be selected from the materials collected at the five quarry locations. This was adjusted in order to more accurately distinguish the disparity in elemental composition from one site to another. The number of samples for the 3GA48 quarry and the Caddo Gap road cut was increased from 4 to 5, and only 2 samples were tested from Oklahoma. The modifications allow for the range of color and texture variation at (3GA48) and Caddo Gap to be properly sampled. The sample areas are located in Hot Springs, Garland and Montgomery counties in central Arkansas. Samples from McCurtain county Oklahoma were included as a control. These samples may help to illustrate the potential variability that might be found throughout the novaculite formations. The majority of the samples were taken from outcrops to ensure an accurate description of the variation in the quarry area as well as within the formation at large. Table 2 lists the total number of samples collected, how many of each were utilized from 32 each site, and whether they came from an outcrop or surface collection. The samples were collected from locations mentioned previously in the geological and archaeological literature. (3GA48) and (3HS603, 3HS69) are in an area of the formation subjected to metamorphism. (Keller et al. 1977; Miser 1943; Sholes 1978). The novaculite outcrops in the Caddo Cap area, according to Sholes (1978) were not altered by metamorphism, although Keller et al. (1977) would disagree. The Caddo Gap outcrop has been extensively studied by geologists (Keller et al. 1977; Sholes 1977,1978; Cornish 1997). Figure 7 shows three photographs of the road cut deposits. The question of metamorphism and its distance from the other locations made this a perfect area for sampling. 33 Table 2: Sample list and provenance: *Not all samples were used for analysis; only those with second sample labels that begin with MBT were tested. Site 3HS69 is an aboriginal quarry on the National Register of Historic Places. Sample Designations Provenance 3GA48 MCQ001-MBT003 Outcrop MCQ002-MBT002 Outcrop MCQ003 Outcrop MCQ004 Outcrop MCQ005-MBT004 Surface, around 004 outcrop MCQ006-MBT005 Surface debris MCQ007-MBT001 Surface near 008 outcrop MCQ008 Outcrop --------------------------------------------------------------------------------------------------------3HS603 & 3HS69 LCT001-MBT008 Outcrop LCT002 Outcrop LCT003 Surface debris LCT004-MBT007 Surface debris LCT005-MBT009 Quarry pit surface debris LCT006-MBT006 Outcrop LCB001-MBT012 Outcrop LCB002-MBT013 Outcrop, same as 002-013 LCB003 Outcrop LCB004-MBT010 Outcrop, south of 004-010 LCB005-MBT011 Surface ---------------------------------------------------------------------------------------------------Caddo Gap Roadcut, 10 total samples CGR001-MBT014 Outcrop CGR002-MBT015 Outcrop CGR003-MBT016 Outcrop CGR004 Outcrop CGR005 Outcrop CGR006-MBT018 Outcrop CGR007 Outcrop CGR008 Outcrop CGR009-MBT017 Outcrop CGR010 Outcrop ----------------------------------------------------------------------------------------------------Oklahoma-Big Hudson Creek Mtn. OK001-MBT019 Surface OK002-MBT020 Surface 34 Fieldwork was conducted over the course of several days, one day for each sample locality, with the exception of 3HS603 and 3HS69 on . Samples were collected from these sites during a one-day quarry survey with the help of Mary Beth Trubitt of the Arkansas Archaeological Survey. Novaculite outcrops on the surface along the slope and top of the mountain ridge. Site 3HS603 is composed of quarry pits and associated debris scatters, and more pits are discovered during every return visit. The quarry pits were dug into the higher quality layers of the formation. The heavily weathered novaculite surface outcrops are white to gray with veins of black and tend to fracture into blocks. This novaculite is less desirable for flint knapping. In contrast, quarry pit debris exhibits a wider variety of colors and is of a higher quality. The 3GA48 novaculite outcrops are extensive, stretching along the entire ridge for several miles. This area has been intensely exploited by prehistoric Native American groups, historic whetstone miners, and modern quarry operators. Prehistoric quarry pits can be found all over the ridge, but most have not been systematically recorded. 3GA48 novaculite comes in a variety of colors and resembles the Indian Mountain novaculite in color as well as texture. The Caddo Gap road cut is located between the towns of Glenwood and Caddo Gap on State Highway 8. It is one of the few places where the entire Arkansas Novaculite formation is represented. This spectacular outcrop highlights the range of variation within the formation. It is also a good example of how the rock layers have been tilted and deformed during the Ouachita orogeny (figure 7). 35 36 This vertical orientation has likely contributed to the practice of pit mining the novaculite to expose specific layers. Direct access to the individual layers of the formation allowed representative samples to be obtained in a few hours. A set of novaculite samples from Big Hudson Creek Mountain in Oklahoma were collected by Meeks Etchieson of the U.S. Forest service. A GPS location was recorded to mark the general location. Chert samples previously analyzed with INAA at MURR were included in the statistical analysis. These samples were originally incorporated because they were thought to be novaculite from the Caballos Formation in western Texas. However, upon further research, these samples were determined to be a variety of chert from the Edwards Formation, which is called “Texas novaculite” by local flint-knappers. This material is not novaculite, but Frederick (personal communication, April 2008) chose to use this name because it had a history of colloquial use in the region. This erroneously named ‘Texas novaculite’ is really a chert inter-bedded with limestone formed during the Cretaceous; significantly younger than novaculite proper. It is called ‘Texas novaculite’ by locals and flint-knappers because it is found nearby the actual Caballos Novaculite outcrops in western Texas. This distinction was not made clear by Frederick et al. (1994) nor was it identified as ‘a variety of Edwards’ in the data provided to MURR when the analysis was initially conducted. Nonetheless, the Texas data were included in the analysis and my discussion of the results under the heading Edwards chert. Photographs were taken at each sample location and GPS coordinates were plotted with a Garmin handheld unit with an accuracy of less than 16 feet. Information about the color, texture, extent of weathering, presence of quarry debris/talus and any 37 other pertinent data was documented for each sample locality. A unique sample label consisting of three letters and three numbers is used to identify them for the chemical testing procedures (table 2). After the samples were collected, they were brought to the Arkansas Archaeological Survey station in Fayetteville for curation. The samples were cleaned and when dry, twenty were selected for the test group following the sample strategy mentioned above. Each sample was broken down into more manageable pieces to facilitate analysis as well as shipping. Small flakes were also collected from each INAA sample for the EDS testing. It was important that a fresh interior surface was created for the EDS testing in particular, as the exterior of the rocks were subjected to physical and chemical erosion which could potentially skew the results. This could also be an issue for later artifact testing in that some manner of artifact destruction even with EDS may be unavoidable. As Luedtke (1978:422) points out: “Artifacts may differ slightly from source materials because of surface alterations occurring after the artifact has been deposited in the soil.” Once the samples were appropriately sized, a photograph was taken of each so that it could be identified and inventoried in the collection database. Each item was bagged with an index card listing its sample number before being sent off to MURR. EDS samples were prepped subsequent to the INAA samples because they require less time for analysis. Because EDS requires test samples to be less than one cubic inch, small flakes were selected from each sample. The flakes were cleaned of surface debris and placed into sealed 4x4 inch 2mil bags. A label was also placed inside to help with identification. The samples were then organized into seven sets, six of which contained 38 three samples and the seventh containing only two. This arrangement was used to decrease the amount of time and money spent for the testing. The samples were then delivered to the ESEM laboratory at the University of Arkansas. All 20 of the samples were first tested at a low power and then samples MBT006 and MBT008, were tested at a slightly higher voltage to get a reading at a deeper level. One sample, MBT007, was tested twice at low power to determine if color differences affected the trace element concentration. A SEM photograph was also taken of each sample. The EDS raw data are in Appendix B. Using the GPS coordinates collected during the fieldwork, maps of the sample locations were created. Several other maps were acquired including a regional map showing general locations of sample areas in Arkansas and Oklahoma as well as the extent of the formation, its outcrops and how they correspond with the topography (figure 8). Comprehensive maps are stored at the Arkansas Archaeological Survey. All sample information was organized to create the collection database, and to facilitate further research. Photographs of the sample locations were assigned a label that corresponds to each sample and a column was added to the spreadsheet to accommodate them. The collection database, in Appendix A, also includes the elemental concentration data in parts per million (ppm) from the INAA. 39 Testing Methodology Instrumental neutron activation analysis (INAA) involves the bombardment of the submitted sample with radiation to determine the elemental composition. This results in radioactive contamination of the samples, which are retained at MURR. Two different irradiations are conducted, a short and a long. Short-lived elements can be tested with little exposure to radiation as they decay faster allowing for a shorter detection period. These short-lived elements “include Al, Ba, Ca, Cl, Dy, K, Mn, Na, Ti, and V” (Glascock et al. 2007). The testing of long-lived elements can take as long as three to four weeks in order to allow proper time for radioactive decay. The samples must be irradiated longer than for the detection of short-lived elements and must be encased in high purity quartz vials as opposed to the polyethylene vials used in the short irradiation. (Boulanger and Glascock 2008). Long-lived elements include As, La, Lu, Nd, Sm, U, Yb, Ce, Co, Cr, Cs, Eu, Fe, Hf, Ni, Rb, Sb, Sc, Ta, Tb, Th, Zn, and Zr. Some of the NAA analyses listed in Luedtke (1992), such as the Arkansas novaculite, were tested at the University of Michigan’s Museum of Anthropology. The following elements were identified: Ba, Br, Ce, Co, Cr, Cs, Eu, Fe, Hf, La, Lu, Na, Rb, Sb, Sc, Sm, Th, U, Yb. She does not describe the methodology at length, but does mention that much of the data relevant to collection and sampling was lost during a relocation of the facility. The instruments that are used to collect the data from the irradiated samples include a semiconductor detector, associated electronics, and a computer-based, multichannel analyzer (Glascock 2007). Dr. Michael Glascock and his staff conducted the INNA at the Missouri University Research Reactor (MURR). 41 Samples were crushed into a powder, which allows for a more accurate collection of trace elements, and subjected to the short and long irradiations. Statistical analyses conducted by the research reactor staff address the elemental decay data. The analysts interpret the data and compile a report that includes statistical analyses and a discussion of the results. Scanning Electron Microscopy-Energy Dispersive X-ray Spectrometry or SEMEDS, is utilized to acquire microscopic views and elemental composition of the surface of the any number of materials. According to Buffalo University’ s South Campus Instrumentation Center’s website, “In scanning electron microscopy (SEM) an electron beam is scanned across a sample's surface. When the electrons strike the sample, a variety of signals are generated, and it is the detection of specific signals, which produces an image or a sample's elemental composition. […] Interaction of the primary beam with atoms in the sample causes shell transitions that result in the emission of an X-ray. The emitted X-ray has an energy characteristic of the parent element. Detection and measurement of the energy permits elemental analysis (Energy Dispersive X-ray Spectroscopy or EDS). EDS can provide rapid qualitative, or with adequate standards, quantitative analysis of elemental composition with a sampling depth of 1-2 microns. X-rays may also be used to form maps or line profiles, showing the elemental distribution in a sample surface” (SCIC website, accessed March 3, 2008). According to Alan Toland, the researcher assigned to run the various chemical and elemental testing equipment for the Physics department, the electron emissions detected by the X-ray detector removes electrons from the outer shells of the individual atoms. The instrument's computer is then able to weigh the amount of each element per sample and take its mass into account in the determination of the final concentration for each element (Toland, personal communication 2008). The EDS at the University of Arkansas involves the use of an SEM and a specialized X-ray element. The EDS element is kept in a liquid nitrogen cooled chamber 42 and must be lowered down into the test chamber manually by the technician. The device is then able to measure the elemental composition of the material and can be directed to pinpoint the samples individually. Major elements can be determined immediately, and trace elements can take several minutes to show up. The longer the machine is left to run the more elements can be detected. EDS can identify a total of 15 elements: C, O, F, Ni, Na, Mg, Al, Is, P, Cl, K, Ca, Mn, Fe, and Zn. Toland estimate’s that it can take about 15 to 20 minutes to get a good reading. Depending on what one is looking for individual testing times can vary from ten to forty-five minutes. (Toland, personal communication 2007). EDS tests whole specimens without destroying or contaminating them. This technique was selected for several reasons. First and foremost the test can be performed relatively quickly and requires little sample preparation. It is also less expensive and is, in principle, non-destructive. This method is compared to INAA in order to determine if it is a suitable alternative. There are however, sample size restrictions. Samples placed in the chamber individually can be about 2.5cm2. Three to four samples can be placed in the chamber together; however, they must be no bigger than 1.0 cm2. Doerr (2004) utilized this same method to characterize local chert and novaculite. The majority of the 36 source and 17 artifact samples were tested two times apiece, resulting in 107 data sets. All samples were tested at the same resolution, but the instrument voltage (10 or 30 KeV), which determines the depth at which the elements are recorded from, is not listed. 43 CHAPTER 4: DATA AND RESULTS 44 EDS (energy dispersive x-ray spectrometry) All of the samples were first analyzed at 10 KeV and SEM photographs were taken. Two samples were also analyzed at 30KeV to determine how much difference there would be between the two tests. A second SEM photo was taken during the 30KeV tests as well. When the instrument is set at 10 KeV, only the surface can be sampled. The higher voltage, 30 KeV, allows for testing beneath the surface. The sample data were generated in two forms, a graphical representation and a numerical readout of the elemental concentrations (see Appendix B). The data collected from the energy dispersive x-ray spectrometry were largely in a graphical format (see example, figure 9). This was converted into elemental concentration data to allow for comparison with the INAA data. 45 All of the samples tested at 10KeV contain Si, O and C. While the highest elemental concentration in most of the samples is Si, a few have higher values for O (MBT001, 005,010). In sample MBT005, from 3GA48 , the C value is also higher than that for Si. Almost all of the samples contain Al in varying concentrations, with the exception of samples MBT010, 011 and 015. Six of the samples contain trace amounts of Mg (MBT001,005, 008,013, 014, and 019). Samples MBT001, 005 and 020 contained trace amounts of P. Trace amounts of Ca are evident in samples MBT001, 005, 008 and 013 and samples MBT008, 013, 014 and 019 contain K. There are also a few samples that have a trace of Na, including MBT002, 007, 009, and 012. Three samples contain trace amounts of the metals Fe and Ti. Sample MBT014 contains a trace amount of Fe and Ti, while MBT008 contains only Ti. Sample MBT007 was tested twice because of its sharp change from a white waxy luster to a dull black. The first test revealed only Si, C and O. The second test of the additional color area revealed Na, Al and S as well as Si, C and O. It is unknown which color is associated with each test. Future EDS testing should include the color/texture change to assess corresponding differences in elemental concentration. This information may have important implications for source identification. Two of the samples, MBT006 and MBT008, were tested a 30Kev to determine how much difference there would be in elemental detection. Si, O, C and Al were present in both tests of sample MBT006. At 30Kev Co, Hg, Fe, Ni, and Sn show up in trace amounts. Sample MBT008 at 30KeV showed Mg, Ba and Fe in addition to the Si, O, Al, S, K and Ca that was evident in the 10KeV test. Several elements that were present in the 10KeV test did not show up at 30KeV. These elements are C, Na, Mg, K, and Ti. 46 It is interesting that there is such a distinct difference between the results of each test voltage. Even though only two samples were tested at 30KeV, it would be prudent to test samples at both voltages in the future to assure that the full range of elements can be obtained. The raw data from the EDS tests of all samples can be found in Appendix B. The EDS data were examined by Matthew Boulanger (personal communication March, 2008) at MURR to determine its fitness for statistical analysis and direct comparison with the INAA data. He identified several problems with the results of the testing, which would preclude comparative efforts. Primarily Si, O, and C, the most common elements identified by the EDS, are not trace elements. They are some of the most common elements on earth and would only be useful if the purpose of the testing was to identify the types of minerals present in the material. Boulanger also points out that inconsistencies in types of elements detected for each sample prevent statistical analysis. Without enough trace elements that are common to all or most of the samples, a direct statistical comparison with the INAA results cannot be undertaken. One outcome of the EDS testing was the identification of major procedural issues with the testing and reporting. A major downfall was the lack of documentation regarding the length of time each sample was in the test chamber. The longer a specimen is in the chamber the better the vacuum created, which increases the amount of trace elements that can be detected. Future testing with this method should involve some amount of experimentation to determine the optimal timing for maximum return. Additionally, it would also be beneficial to test all samples at 30Kev to further ensure that all possible trace elements are detected. 47 INAA (instrumental neutron activation analysis) Instrumental neutron activation analysis (INAA) involves the crushing of the sample into a powder, which allows testing of the entire sample resulting in a higher detection rate for trace elements. This provides a better characterization of the material. This analysis was conducted by the staff at the University of Missouri’s research reactor (MURR), who wrote a comprehensive report of the results. The report includes the raw data on an excel spreadsheet, maps, charts and graphs of the statistical analyses and a written interpretation of the results. The element concentration data are reported in parts per million and tabulated with Microsoft Office Excel (Boulanger and Glascock 2008). Because of the small sample size, data from a previous study conducted at MURR with samples of Edwards chert from Texas (Fredrick et al. 1994) were included in the statistical analyses. Larger sample sizes are required for many statistical procedures such as principle component analysis utilized herein (Boulanger and Glascock 2008). Other statistical procedures applied to the data include hierarchical cluster analysis (HCA) and elemental bivariate plots. The Texas data were also included in some of these analyses. A copy of the original report from MURR is included as Appendix C. This report does not reflect the recognition that the Texas material is chert, not novaculite as it was made before the issue was identified. During an examination of the raw data, several observations were made. All of the samples contain the following elements: Ba, La, Lu, Nd, Sm, U, Ce, Co, Cr, Cs, Eu, Fe, Hf, Rb, Sb, Sc, Ta, Tb, Th, Zn, Al, Dy, Mn, and Na. All but one sample, MBT016, contain Sr. High concentrations of Al are evident in all of the samples. High concentrations of K and Fe are also present in the samples with the exception of 48 MBT010, which contains no K at all. Ba was also present in moderate concentrations; most of the values are over 10 ppm with the exception of MBT008 and MBT010 (5.5497 and 3.8687 respectively). Bivariate plots of the principle component results (figure 10 and 11) show a distinct disparity between the Arkansas-Oklahoma samples and those from Texas (Fredrick et al. 1994). Three other statistical populations can been seen in the PCA bivariate plots which correspond with 1) in (3HS603, 3HS69); 2) Caddo Gap and; 3) (3GA48). The Oklahoma samples cluster together as well and further testing may prove to separate them from the Arkansas materials. These groups, while based on a small number of samples, show a great deal of promise. Similar groupings of the sample areas can be seen in the elemental bivariate plots. The U and Th concentrations from Arkansas-Oklahoma novaculite, Edwards chert and Florence A chert from Oklahoma, were compared resulting in the bivariate plot in figure 12. The results show that the Arkansas-Oklahoma samples, with the exception of MBT015, are low in U but have higher concentrations of Th (Boulanger and Glascock 2008). The Edwards chert is similar to the Florence A chert in its higher U levels. Similar results are seen in the Eu and U bivariate plot in figure13, which compares the Arkansas-Oklahoma novaculite with all of the cryptocrystalline silicate rocks ever analyzed by MURR from Oklahoma and Texas. One interesting thing about this particular diagram is that one of the samples, an artifact from the previously analyzed materials, falls into the same area of the graph as the Arkansas-Oklahoma novaculite. Results of the Cr and Mn bivariate plot are similar to the PCA plots in figures 10 and 11 49 in that the three groups correspond to (3HS603, 3HS69), (3GA48) and Caddo Gap (figure 14). Hierarchical cluster analysis (HCA) also reinforces the disparity between the Frederick et al. (1994) samples from Texas and those from Arkansas and Oklahoma tested for this study. Figure 15 shows the HCA chart. Sample MBT007 (from 3HS603/ was found to be closer to the Texas samples based on mean Euclidean distance, but on its own separate branch of the cluster. It also stands apart when plotted with the Arkansas-Oklahoma samples alone (figure 16). The exact nature of the difference between MBT007 and the rest of the samples is uncertain. What is most interesting is that it was also identified as an anomaly during the EDS testing. The sample has two distinct colorations, white and black. The black can be seen intruding into the white as small thin strands. The colors also correspond with different textures, which may have some relationship to the particular trace elements present within the material. This sample was collected from the talus debris pile that surrounded a large quarry pit. MBT007 resembles the novaculite from outcrops in the area and may have been excavated from the higher quality, un-weathered veins buried beneath the topsoil. It is highly unlikely that the sample was transported from another source. The small number of samples did present a problem when it came to applying statistical evaluation to the data. Nevertheless, a few general observations can be made which should help to inform future novaculite characterization studies. The analyses conducted by Boulanger and Glascock (2008:9) demonstrate that Arkansas novaculite is “chemically heterogeneous and quite variable.” None of the previous characterization studies involving novaculite made any interpretations regarding its nature and 50 composition, other than to report the statistical data (Ives 1984; Luedtke 1992; Doerr 2004). Based on the INAA results from this study, it is possible to distinguish between the Arkansas-Oklahoma novaculite and all other cryptocrystalline silicates previously analyzed by MURR from Texas and Oklahoma. Without further analysis of a larger number of samples, it is not possible to determine if individual quarry sites can be identified with this method. 51 3GA48 3HS69/603 3GA48 3HS69/603 54 55 3GA48 3HS69/603 3GA48 3HS69/603 3GA48 3HS69/603 CHAPTER 5: CONCLUDING REMARKS 59 The purpose of this study was to determine whether the Arkansas novaculite was heterogeneous enough to warrant further characterization activities and to ascertain whether EDS would be a useful alternative to destructive INAA testing. Based on the results of the study several conclusions were made to address the questions posed early on in the research. 1) Can we detect intra-formational variation in Novaculite? The short answer is yes, there is definitely a discernible amount of variation throughout the extent of the novaculite deposits in Arkansas and Oklahoma. The formation contains a significant amount of heterogeneity based on the elemental data. Neutron activation analysis identified several statistical populations within the data that are generally correlated with the sample locations. These groupings have the potential to become more distinct with further testing of a larger more representative sample. Future studies should also include the Caballos Novaculite in Texas. 2) If there is variation, is it significant? The disparity that was identified between the Arkansas-Oklahoma samples in the statistical analysis shows that a potential for significant variation does exist. The only way to determine this with certainty would be to conduct a larger characterization study, which will involve more intensive sampling of the area to the west of Caddo Gap. It would also be essential to sample novaculite outcrops and quarries in Oklahoma and Texas. Boulanger and Glascock point out that based on the small number of samples the 60 variation within each sample location is greater than the whole. This can only be clarified with a larger more representative sample size. 3) Is the variation enough to differentiate between source areas, especially those in close proximity? Using statistical analyses, such as principle component and hierarchal cluster analysis, several groups were identified. The samples tended to cluster into groups that corresponded to their sample locations. While there was a significant amount of variation within the novaculite investigated in this study, the sample size was too small to definitively identify distinct source areas. However, the several statistical populations that were identified in the INAA are encouraging and provide a glimpse into the potential for individual quarry identification. 4) Can we use these data to identify the particular quarries as raw materials sources for artifact production? Will we be able to determine the source of individual artifacts? As mentioned, the presence of statistical groupings in the data tends to identify the samples that correlated with those from the same sample location. At present, the data from both INAA and EDS are insufficient to identify artifactual source areas with any confidence. Fortunately, the data from this study can be utilized in subsequent research. Further analysis on a larger population of samples must be conducted before any definitive statements can be made. 5) Is there a difference between the results from INAA and EDX? If so what is the difference and what are the implications for future research? 61 INAA has long been the standard in archaeological trace element identification. It is reliable, well tested and remains at the forefront of characterization methodology. The destructive nature of this technique is often an undesirable outcome especially with relation to artifact testing. In contrast, EDS does not require the destruction of materials to determine their elemental composition. This advantage over INAA is significant, but it may be the only advantage. The EDS instrument utilized for this study tests the surface of an object at low voltage and just below at a higher voltage. Based on the interpretation of the data recovered from both EDS tests, surface leaching may have a significant impact on the ability to detect trace elements with this technique. The materials tested at a higher voltage did detect more trace elements, but only two samples were tested in this manner. Further testing should be conducted to determine if the higher voltage makes a difference in the results. The suite of elements identified by EDS are not the same as those identified by INAA. This has a significant impact on the results in that elements not shared by both techniques cannot be included in statistical analyses. These incompatible elements may be significant in determining the difference between outcrops or sources, but would be ignored by the analysis. Another issue with EDS is the lack of a consistent testing methodology. More testing needs to be carried out in order to establish parameters that will provide reliable and comparable results. That being said, the two techniques have produced results that are encouraging for future research into this topic. However, INAA is a more reliable well-tested approach for trace element characterization. EDS may still be a useful technique, but further and more intensive testing must occur before it can be utilized as an alternative to INAA. 62 Based on these results a full-scale characterization study of the Arkansas Novaculite is recommended. An in-depth elemental characterization of the intraformational variation will provide plenty of data for raw material identification and artifact comparison in the future. It would also be prudent to research and characterize the Caballos Novaculite in the Marathon region of Texas. Further, excavations in and around the quarry pits at 3HS603 in should be conducted in order to identify the character of the novaculite that was quarried from the pits. This would provide a more in-depth characterization of the formation as well as a better understanding of prehistoric raw material preferences. The addition of these data are likely to allow the sourcing of novaculite artifacts and provide the archaeological community with a comprehensive database for comparison. 63 REFRENCES CITED Aber, James S. 2003 Ouachita Mountains. Electronic document, http://academic.emporia.edu/aberjame/struc_geo/ouachita/ouachita.htm, accessed March 3, 2008. Acquafredda, P., T. Andriani, S. Lorenzoni and E. Zanettin 1999 Chemical Characterization of Obsidians from Different Mediterranean Sources by Non-destructive SEM-EDS Analytical Method. Journal of Archaeological Science 26: 315-325. Ambroz, Jessica A., Michael D. Glascock, and Craig E. Skinner. 2001 Chemical Differentiation of Obsidian Within the Glass Buttes Complex, Oregon. Journal of Archaeological Science 28(7): 741-746 Baker, Charles M. 1974 A Study of Aboriginal Novaculite Exploitation in the Ouachita Mountains of South-Central Arkansas. Unpublished Master’s thesis, Department of Anthropology, University of Arkansas, Fayetteville. Blackman, M.J. and R.L. Bishop 2007 The Smithsonian-NIST Partnership: The Application of Instrumental Neutron Activation Analysis to Archaeology. Archaeometry 49(2): 321-341. Boulanger, Matthew T. and Michael D. Glascock 2008 Neutron Activation Analysis of Novaculite from Garland, Montgomery, and Hot Springs Counties, Arkansas and from McCurtain County, Oklahoma. On file at the Archaeometry Laboratory, University of Missouri Research Reactor, Columbia, MO. Church, Tim., Julie E. Francis and Cherie E. Haury 1994 Lithic Resource Studies: A Sourcebook for Archaeologists. Special Publication, Lithic Technology. Department of Anthropology, University of Tulsa. Cornish, Christopher, S. 1997 Trace Element and Isotope Geo-chemistry of the Laminated Facies of the Lower Member of the Arkansas Novaculite, Benton Uplift, Arkansas. Unpublished Masters thesis, Department of Geology, Stephen F. Austin State University, Texas. 64 Doerr, Erica. 2004 The Silica Source for the Mississippian Chert and its Possible Elemental Signature in Northern Arkansas and Southwestern Missouri. Unpublished Master’s thesis, Department of Anthropology, University of Arkansas, Fayetteville. Ericson, Jonathon E. and Barbara A. Purdy 1984 Prehistoric Quarries and Lithic Production. Cambridge University Press, New York. Etchieson, M. 1997 Prehistoric Novaculite Quarries in the Ouachita Mountains. Paper presented at the Annual Meeting of the Society for American Archaeology, Nashville, IN. Web published at http://www. fs. fed. us/r8/Ouachita/publications/prehistoric novaculite quarries.pdf Featherstonhaugh, G.W. 1968 [1844] Excursion through the Slave States, from Washington on the Potomac to the Frontier of Mexico; with Sketches of Popular Manners and Geological Notices. Harper & Brothers. 1968 reprint ed. Negro Universities Press, Greenwood Publishing , New York. Flenniken J. Jeffery and Ervan G. Garrison 1975 Thermally Altered Novaculite and Stone Tool Manufacturing Techniques Journal of Field Archaeology, 2.5: 125-131 Folk, Robert L. and Earle F. McBride 1976 The Caballos Novcaulite Revisited Part I: Origin of Novaculite Members. Journal of Sedimentary Petrology 46(3): 659-669. Folk, Robert L. and Earle F. McBride 1977 The Caballos Novaculite Revisited Part II: Chert and Shale Members and Synthesis. Journal of Sedimentary Petrology 47(3): 1261-1286. Frederick, Charles D., Michael D. Glascock, Hector Neff and Christopher M. Stevenson 1994 Evaluation of Chert Patination as a Dating Technique: A Case Study from Fort Hood, Texas. Research Report No. 32, Archaeological Resource Management Series. United States Army, Fort Hood, TX. Glascock, M.D., R.J. Speakman and H. Neff 2007 Archaeometry at the University of Missouri Research Reactor and the Provenance of Obsidian Artifacts in North America. Archaeometry 49(2): 343357. 65 Glascock, Michael. D 2007 Overview of Neutron Activation Analysis. Electronic document, http://archaeometry.missouri.edu/naa overview.html, accessed November 28th, 2007. Goldstein A. Jr. and T.A. Hendricks 1953 Siliceous Sediments of Ouachita Facies in Oklahoma. Bulletin of the Geological Society of America, 64(4): 421-442. Griswold, L. S. 1892 Whetstones and the Novaculites of Arkansas. Volume III of Annual Report of the Geological Survey of Arkansas for 1890, by John C. Branner. Press Printing Co., Little Rock, AR. Grunig, Diana 1977 The Geologic History of the Austin Area. In Guidebook of the Geology of Travis County. University of Texas, electronic document. www.lib.utexas.edu/geo/ggtc/toc.html, accessed March 3, 2008. Hancock, R.G.V., L.A. Pavish, S. Aufreiter 2007 Archaeometry at Slowpoke-Toronto. Archaeometry 49(2): 229-243. Hoard, R. J., J. R. Bozell, S. R. Holen, M. D. Glascock, H Neff, and J. M. Elam 1993 Source Determination of White River Group Silicates from Two Archaeological Sites in the Great Plains. American Antiquity 58(4): 698-710. Holbrook, Drew. F. and Charles. G. Stone 1979 Arkansas Novaculite - A Silica Resource. Arkansas Geological Commission, Little Rock. Holmes, W. H. 1891 Aboriginal Novaculite Quarries in Garland County, Arkansas. American Anthropologist 4 (old series): 313-316. Holmes, W. H 1892 Modem Quarry Refuse and the Paleolithic Theory. Science 20(512):295297. Holmes, W. H 1919 [1974] Handbook of Aboriginal American Antiquities, Part I Introductory, The Lithic Industries. Smithsonian Institution, Bureau of American Ethnology Bulletin 60, Washington, D.C., reprinted 1974, Burt Franklin, New York, NY. 66 Honess, C.W. 1923 Geology of the Southern Ouachita Mountains of Oklahoma, Part I: Stratigraphy, Structure, and Physiographic History. Oklahoma Geological Survey, Bulletin 32: 108-139, Norman, Ok. Ives, D. J. 1984 Neutron Activation Analysis Characterization of Selected Prehistoric Chert Quarrying Areas. Ph.D. dissertation, Department of Anthropology, University of Missouri-Columbia. Jenney, W. P. 1891 Ancient Novaculite Mines near Magnet Cove, Hot Springs County, Arkansas. American Anthropologist 4 (old series):316-318. Julig, P. J. 1995 The Sourcing of Chert Artifacts by INAA: Some Examples from the Great Lakes Region. Journal of World Anthropology Vol. 1, NO.2. Electronic journal article available at http://wings.buffalo.edu/rcsearch/anthrogis/JWA/VIN2/juligpap.html. Keller, W.D., George W. Viele and Clayton H. Johnson 1977 Texture of Arkansas Novaculite Indicates Thermally Induced Metamorphism. Journal of Sedimentary Petrology 47(2): 834-843. Keller, W. D., C. G. Stone, and A. L. Hoersch 1985 Textures of Paleozoic Chert and Novaculite in the Ouachita Mountains of Arkansas and Oklahoma and Their Geological Significance. Geological Society of America Bulletin 96: 1353-1363. Lemley, Harry J. 1962 Prehistoric Novaculite Quarries of Arkansas. Arkansas Archaeologist, Arkansas Archaelogical Survey 3(2): 13-15. Lowe, Donald R. 1977 The Arkansas Novaculite: Some Aspects of its Physical Sedimentation. In Symposium on the Geology of the Ouachita Mountains: Stratigraphy, Sedimentology, Petrography, Tectonics, and Paleontology, edited by C.G. Stone, Vol. 1:pp. 139-145. Miscellaneous Publication 13 of the Arkansas Geological Survey. Luedtke, B. E. 1978 Chert Sources and Trace-Element Analysis. American Antiquity 43(3):413423. 67 Luedtke, B. E. 1992 An Archaeologist's Guide to Chert and Flint. Archaeological Research Tools No. 7, Institute of Archaeology, University of California, Los Angeles. McFarland, John D. 2004 Stratigraphic Summary of Arkansas. Information Circular 36, revised. Arkansas Geological Commission, Little Rock. Miser, Hugh.D. and A.H. Purdue 1929 Geology of the DeQueen and Caddo Gap Quadrangles, Arkansas. U.S. Geological Survey, Bulletin 808. Miser, Hugh D. 1943 Quartz Veins in the Ouachita Mountains of Arkansas and Oklahoma, their Relations to Structure, Metamorphism, and Metalliferous Deposits Economic Geology, 38: 91-118. Niem, A.R. 1977 Mississippian pyroclastic flow and ash-fall deposits in the deep-marine Ouachita flysch basin, Oklahoma and Arkansas. Geological Society of America Bulletin, 88(1): 49-61. Owen, David D. 1860 Second Report of a Geological Reconnaissance of the Middle and SouthernCounties of Arkansas: made during the years 1859 and 1860. C. Sherman and Son Printers, Philadelphia. South Campus Instrumentation Center at Unicersity of Buffalo, NY. 2008 SEM/EDS : Scanning Electron Microscopy with X-ray microanalysis. Electronic document, http://www.sdm.buffalo.edu/scic/sem-eds.html, accessed March 3, 2008. Sholes, Mark Allen 1977 Arkansas Novaculite Stratigraphy. In Symposium on the Geology of the Ouachita Mountians, edited by C.G.Stone, pp. 139-145. Miscellaneous Publication 13 of the Arkansas Geological Survey. Sholes, Mark Allen 1978 Stratigraphy and Petrography of the Arkansas Novaculite of Arkansas and Oklahoma. Unpublished PhD. dissertation, Department of Geosciences, University of Texas, Austin. Speakman, R.L. and M.D. Glascock 2007 Special Issue: Acknowledging Fifty Years of Neutron Activation Analysis in Archaeology. Archaeometry 49(2): 179-183. 68 Trubitt, Mary Beth. 2005a Mapping a Novaculite Quarry in Hot Springs National Park. Caddoan Archeology Journal 14:17-33. Trubitt, Mary Beth 2005b Understanding the Organization of Novaculite Tool Production. Paper presented in "Lithic Reduction Analysis and Problems in Prehistory" Symposium, 70th Annual Meeting of the Society for American Archaeology, Salt Lake City, Utah . Trubitt, Mary Beth, Tom Green, and Ann Early 2004 A Research Design for Investigating Novaculite Quarry Sites in the Ouachita Mountains. The Arkansas Archeologist (Bulletin of the Arkansas Archeological Society) 43(for 2002): 17-62. 69 Appendix A: Novaculite Collection Database 70 71 74 75 76 Appendix B: EDS Raw Data 77 78 79 80 81 82 83 84 Appendix C: INAA Report from MURR 85 Neutron Activation Analysis of Novaculite from Garland, Montgomery, and Hot Spring Counties, Arkansas, and from McCurtain County, Oklahoma Prepared by: Matthew T. Boulanger and Michael D. Glascock Archaeometry Laboratory, University of Missouri Research Reactor Columbia, MO 65211 Prepared for: Ms. Kristin Scarr Department of Anthropology University of Arkansas Arkansas Archaeological Survey 2475 N. Hatch Avenue Fayetteville, AR 72704 March 4, 2008 86 3GA48 3HS69/603 irradiations at MURR. At the same time, 800 mg aliquots from each sample were weighed into clean high-purity quartz vials used for long irradiations. Individual sample weights were recorded to the nearest 0.01 mg using an analytical balance. Both vials were sealed prior to irradiation. Along with the unknown samples, standards made from National Institute of Standards and Technology (NIST) certified standard reference materials of SRM-1633a (Coal Fly Ash), SRM-278 (Obsidian Rock), and SRM-688 (Basalt Rock) were similarly prepared. Irradiation and Gamma-Ray Spectroscopy Neutron activation analysis of most archaeological samples at MURR, which consists of two irradiations and a total of three gamma counts, constitutes a superset of the procedures used at most other NAA laboratories (Glascock 1992; Glascock and Neff 2003; Neff 2000). As discussed in detail by Glascock (1992), a short irradiation is carried out through the pneumatic tube irradiation system. Samples in the polyvials are sequentially irradiated, two at a time, for five seconds by a neutron flux of 8 x 1013 n cm-2 s-1 The 720-second count yields gamma spectra containing peaks for nine short-lived elements aluminum (Al), barium (Ba), calcium (Ca), dysprosium (Dy), potassium (K), manganese (Mn), sodium (Na), titanium (Ti), and vanadium (V). The long-irradiation samples are encapsulated in quartz vials and are subjected to a 70–hour irradiation at a neutron flux of 5 x 1013 n cm-2 s-1. This long irradiation is analogous to the single irradiation utilized at most other laboratories. After the long irradiation, samples decay for seven days, and then are counted for 1800 seconds (the "middle count") on a high-resolution germanium detector coupled to an automatic sample changer. The middle count yields determinations of seven medium half-life elements, namely arsenic (As), lanthanum (La), lutetium (Lu), neodymium (Nd), samarium (Sm), uranium (U), and ytterbium (Yb). After an additional two- or three-week decay, a final count of 8500 seconds is carried out on each sample. The latter measurement yields the following 17 long half-life elements: cerium (Ce), cobalt (Co), chromium (Cr), cesium (Cs), europium (Eu), iron (Fe), hafnium (Hf), nickel (Ni), rubidium (Rb), antimony (Sb), scandium (Sc), strontium (Sr), tantalum (Ta), terbium (Tb), thorium (Th), zinc (Zn), and zirconium (Zr). The element concentration data from the three measurements are tabulated in parts per million using Microsoft® Office Excel. Descriptive data for archaeological samples were appended to the concentration spreadsheet. The data are also stored in a dBase/FoxPro database file useful for organizing, sorting, and extracting sample information. Interpreting Chemical Data Analyses at MURR described previously produce elemental concentration values for 32 elements in most analyzed rock samples. However, cryptocrystalline silicates do not always have sufficient quantities of these 32 elements to be detectable using the above procedures. Samples were segregated into groups based upon the locations from which they were collected. Any element not present in greater than 50% of the samples comprising each group was eliminated from this analysis. This procedure eliminated the following elements: Ti, Ca, and Ni. Statistical analyses were subsequently carried out on base-10 logarithms of concentrations on the remaining 29 elements. Use of log concentrations rather than raw data compensates for differences in magnitude between the major elements, such as sodium, and trace elements, such 88 as the rare earth or lanthanide elements (REEs). Transformation to base-10 logarithms also yields a more normal distribution for many trace elements. The interpretation of compositional data obtained from the analysis of archaeological materials is discussed in detail elsewhere (e.g., Baxter and Buck 2000; Bieber et al. 1976; Bishop and Neff 1989; Glascock 1992; Harbottle 1976; Neff 2000) and will only be summarized here. The main goal of data analysis is to identify distinct homogeneous groups within the analytical database. Based on the provenance postulate of Weigand et al.(1977), different chemical groups may be assumed to represent geographically restricted sources. For lithic materials such as obsidian, basalt, and cryptocrystalline silicates (e.g., chert, flint, or jasper), raw material samples are frequently collected from known outcrops or secondary deposits and the compositional data obtained on the samples is used to define the source localities or boundaries. The locations of sources can also be inferred by comparing unknown specimens (i.e., ceramic artifacts) to knowns (i.e., clay samples) or by indirect methods such as the “criterion of abundance” (Bishop et al. 1982) or by arguments based on geological and sedimentological characteristics (e.g., Steponaitis et al. 1996). The ubiquity of ceramic raw materials usually makes it impossible to sample all potential “sources” intensively enough to create groups of knowns to which unknowns can be compared. Lithic sources tend to be more localized and compositionally homogeneous in the case of obsidian or compositionally heterogeneous as is the case for most cherts. Compositional groups can be viewed as “centers of mass” in the compositional hyperspace described by the measured elemental data. Groups are characterized by the locations of their centroids and the unique relationships (i.e., correlations) between the elements. Decisions about whether to assign a specimen to a particular compositional group are based on the overall probability that the measured concentrations for the specimen could have been obtained from that group. Initial hypotheses about source-related subgroups in the compositional data can be derived from non-compositional information (e.g., archaeological context, decorative attributes, etc.) or from application of various pattern-recognition techniques to the multivariate chemical data. Some of the pattern recognition techniques that have been used to investigate archaeological data sets are cluster analysis (CA), principal components analysis (PCA), and discriminant analysis (DA). Each of the techniques has its own advantages and disadvantages which may depend upon the types and quantity of data available for interpretation. The variables (measured elements) in archaeological and geological data sets are often correlated and frequently large in number. This makes handling and interpreting patterns within the data difficult. Therefore, it is often useful to transform the original variables into a smaller set of uncorrelated variables in order to make data interpretation easier. Of the above-mentioned pattern recognition techniques, PCA transforms the data from the original correlated variables into uncorrelated variables most easily. Principal components analysis creates a new set of reference axes arranged in decreasing order of variance subsumed. The individual PCs are linear combinations of the original variables. The data can be displayed on combinations of the new axes, just as they can be displayed on the original elemental concentration axes. PCA can be used in a pure pattern-recognition mode, i.e., to search for subgroups in an undifferentiated data set, or in a more evaluative mode, i.e., to 89 assess the coherence of hypothetical groups suggested by other criteria. Generally, compositional differences between specimens can be expected to be larger for specimens in different groups than for specimens in the same group, and this implies that groups should be detectable as distinct areas of high point density on plots of the first few components. Principal components analysis of chemical data is scale dependent, and analyses tend to be dominated by those elements or isotopes for which the concentrations are relatively large. As a result, standardization methods are common to most statistical packages. A common approach is to transform the data into logarithms (e.g., base 10). As an initial step in the PCA of most chemical data at MURR, the data are transformed into log concentrations to equalize the differences in variance between the major elements such as Al, Ca and Fe, on one hand and trace elements, such as the rare-earth elements (REEs), on the other hand. An additional advantage of the transformation is that it appears to produce more nearly normal distributions for the trace elements. One frequently exploited strength of PCA, discussed by Baxter (1992), Baxter and Buck (2000), and Neff (1994; 2000), is that it can be applied as a simultaneous R- and Q-mode technique, with both variables (elements) and objects (individual analyzed samples) displayed on the same set of principal component reference axes. A plot using the first two principal components as axes is usually the best possible two-dimensional representation of the correlation or variancecovariance structure within the data set. Small angles between the vectors from the origin to variable coordinates indicate strong positive correlation; angles at 90 degrees indicate no correlation; and angles close to 180 degrees indicate strong negative correlation. Likewise, a plot of sample coordinates on these same axes will be the best two-dimensional representation of Euclidean relations among the samples in log-concentration space (if the PCA was based on the variance-covariance matrix) or standardized log-concentration space (if the PCA was based on the correlation matrix). Displaying both objects and variables on the same plot makes it possible to observe the contributions of specific elements to group separation and to the distinctive shapes of the various groups. Such a plot is commonly referred to as a “biplot” in reference to the simultaneous plotting of objects and variables. The variable inter-relationships inferred from a biplot can be verified directly by inspecting bivariate elemental concentration plots. Whether a group can be discriminated easily from other groups can be evaluated visually in two dimensions or statistically in multiple dimensions. A metric known as the Mahalanobis distance (or generalized distance) makes it possible to describe the separation between groups or between individual samples and groups on multiple dimensions. The Mahalanobis distance of a specimen from a group centroid (Bieber et al. 1976; Bishop and Neff 1989) is defined by: Dy2, X = [ y − X ]t I x [ y − X ] where y is the 1 x m array of logged elemental concentrations for the specimen of interest, X is the n x m data matrix of logged concentrations for the group to which the point is being compared with X being it 1 x m centroid, and Ix is the inverse of the m x m variance-covariance matrix of group X. Because Mahalanobis distance takes into account variances and covariances in the multivariate group it is analogous to expressing distance from a univariate mean in standard deviation units. Like standard deviation units, Mahalanobis distances can be converted 90 into probabilities of group membership for individual specimens. For relatively small sample sizes, it is appropriate to base probabilities on Hotelling’s T2, which is the multivariate extension of the univariate Student’s t. When group sizes are small, Mahalanobis distance-based probabilities can fluctuate dramatically depending upon whether or not each specimen is assumed to be a member of the group to which it is being compared. Harbottle (1976) calls this phenomenon “stretchability” in reference to the tendency of an included specimen to stretch the group in the direction of its own location in elemental concentration space. This problem can be circumvented by cross-validation, that is, by removing each specimen from its presumed group before calculating its own probability of membership (Baxter 1994; Leese and Main 1994). This is a conservative approach to group evaluation that may sometimes exclude true group members. Small sample and group sizes place further constraints on the use of Mahalanobis distance: with more elements than samples, the group variance-covariance matrix is singular thus rendering calculation of Ix (and D2 itself) impossible. Therefore, the dimensionality of the groups must somehow be reduced. One approach would be to eliminate elements considered irrelevant or redundant. The problem with this approach is that the investigator’s preconceptions about which elements should be discriminate may not be valid. It also squanders the main advantage of multielement analysis, namely the capability to measure a large number of elements. An alternative approach is to calculate Mahalanobis distances with the scores on principal components extracted from the variance-covariance or correlation matrix for the complete data set. This approach entails only the assumption, entirely reasonable in light of the above discussion of PCA, that most group-separating differences should be visible on the first several PCs. Unless a data set is extremely complex, containing numerous distinct groups, using enough components to subsume at least 90% of the total variance in the data can be generally assumed to yield Mahalanobis distances that approximate Mahalanobis distances in full elemental concentration space. Lastly, Mahalanobis distance calculations are also quite useful for handling missing data (Sayre 1975). When many specimens are analyzed for a large number of elements, it is almost certain that a few element concentrations will be missed for some of the specimens. This occurs most frequently when the concentration for an element is near the detection limit. Rather than eliminate the specimen or the element from consideration, it is possible to substitute a missing value by replacing it with a value that minimizes the Mahalanobis distance for the specimen from the group centroid. Thus, those few specimens which are missing a single concentration value can still be used in group calculations. Results and Discussion The NAA results were entered into a spreadsheet and combined with the provided descriptive data to create a database for sorting and extracting of quarry subgroups. Summary statistics for each source locality are provided in Appendix A, and the combined chemical and descriptive data are provided in Appendix B. A Microsoft Excel document is also provided containing the chemical and descriptive data. As stated above, only 29 of a possible 32 elements were present in sufficient amounts in all samples to be useful for multivariate analysis. However, PCA requires at least two more 91 observations (samples) than the total number of variables (elements). Therefore, previously acquired geochemical data from a novaculite source in Texas (Frederick et al. 1994) were merged with these new data in order to conduct an RQ-mode PCA. Merging these data required the removal of an additional five elements (As, Lu, Dy, K, and V) not present in detectable quantities in the Texas samples. Table 2 lists the eigenvalues and percentages of variance explained by each of the eigenvectors in the PCA. The PCA demonstrates that greater than 90% of the cumulative variance in the dataset is subsumed by the first five principal components (PCs). The first eigenvector is negatively loaded on REEs including Yb, Eu, and Tb, whereas the second eigenvector is loaded with transition metals Co and Mn. A biplot of the first two PCs reveals that samples from Arkansas and Oklahoma, while not easily distinguished from each other, may be separated from Texas novaculite quite readily (Error! Reference source not found. and Error! Reference source not found.). Hierarchical cluster analysis (HCA) using mean Euclidean distance further demonstrates a somewhat high degree of heterogeneity among novaculite sources in Arkansas and Oklahoma (Error! Reference source not found.), yet clear differentiation between these samples and those from Texas (Error! Reference source not found.). Of note is that sample MBT007 from is consistently identified as an outlier using HCA, and in some ways is more similar to Texas samples than to those from Arkansas and Oklahoma. Elemental bivariate plots of the chemical data further support the conclusion that the Arkansas and Oklahoma novaculite samples can be clearly distinguished from regional chert sources (Error! Reference source not found.). Further, comparison of the novaculite data against all the MURR Texas-chert database (405 samples from over 25 different sources, and 309 artifacts) demonstrates that the novaculite sources of Oklahoma and Arkansas may be identified by a low uranium and relatively elevated REE compositional profile (Error! Reference source not found.). However, intra-source chemical variability in the novaculite samples obscures any inter-source differences (Error! Reference source not found.). Conclusions Given the small number of samples representing each source locality, these results must be viewed as preliminary and our interpretations seen as somewhat qualitative. The extremely small number of samples analyzed precludes our ability to use rigorous statistical evaluation techniques such as jack-knifed Mahalanobis distance calculations or canonical discriminant functions; however, some general statements can be made based upon these data. These analyses demonstrate that novaculite from Arkansas and Oklahoma is chemically heterogeneous and quite variable. Intra-source variation is significant, and obscures inter-source chemical differences. That is, the variation in chemistry at any single source is as great or greater than the variation between sources. These results suggest that it will likely be difficult to confidently discriminate among specific source areas, or to confidently assign archaeological artifacts to a single novaculite source within the Ouachita Mountains using NAA alone. Admittedly, it is possible that by collecting samples from archaeological contexts, pieces that are not derived from locally available stone may have been inadvertently collected. However, as there is currently no independent means of assessing whether this is the case, aside from 92 collecting stone directly from in situ geological contexts, it is difficult to evaluate this proposition without more-rigorous sample collection and testing. These analyses have demonstrated that clear chemical differences exist between Texas novaculite and novaculite originating in Arkansas and Oklahoma. It is also possible to clearly distinguish Arkansas and Oklahoma novaculite from virtually all other Texas- and Oklahomaderived cryptocrystalline silicates so far analyzed at MURR. Given these results, investigations into broad-scale long-distance transport of Arkansas novaculite may be worthwhile. Acknowledgements Wesley G. Stoner and Corinne Rosania were responsible for sample preparation and analysis of these samples by INAA. This project was supported in part by NSF grant BCS-0504015 to the Archaeometry Laboratory of the Research Reactor, University of Missouri. Any errors in interpretation are the responsibility of the authors. 93 References Baxter, Mike J. 1992 Archaeological Uses of the Biplot—A Neglected Technique? In Computer Applications and Quantitative Methods in Archaeology, 1991, edited by Gary Lock and Jonathon Moffett, pp. 141-148. BAR International Series. vol. S577. Tempvs Reparatvm, Oxford. 1994 Exploratory Multivariate Analysis in Archaeology. Edinburgh University Press, Edinburgh. Baxter, Mike J. and Caitlin E. Buck 2000 Data Handling and Statistical Analysis. In Modern Analytical Methods in Art and Archaeology, edited by Enrico Ciliberto and Giuseppe Spoto, pp. 681746. John Wiley and Sons, New York. Bieber, Alan M. Jr., Dorothea W. Brooks, Garman Harbottle and Edward V. Sayre 1976 Application of Multivariate Techniques to Analytical Data on Aegean Ceramics. Archaeometry 18:59-74. Bishop, Ronald L. and Hector Neff 1989 Compositional Data Analysis in Archaeology. In Archaeological Chemistry IV, edited by R. O. Allen, pp. 576-586. Advances in Chemistry. vol. 220. American Chemical Society, Washington, D.C. Bishop, Ronald L., Robert L. Rands and George R. Holley 1982 Ceramic Compositional Analysis in Archaeological Perspective. Advances in Archaeological Method and Theory 5:275-330. Boulanger, Matthew T., Allen D. Hathaway, Robert J. Speakman and Michael D. Glascock 2005 A Preliminary Study on the Suitability of Instrumental Neutron Activation Analysis (INAA) for Identifying Hathaway Formation Chert from the Northern Champlain Valley of Vermont. Archaeology of Eastern North America 33:105126. Elam, J. Michael and Michael D. Glascock 1989 Report for Project #1354: Source Area Definition for Florence-A (Kay County) Chert. Archaeometry Laboratory, University of Missouri Research Reactor. Submitted to S. C. Vehik, University of Oklahoma. Frederick, Charles D., Michael D. Glascock, Hector Neff and Christopher M. Stevenson 1994 Evaluation of Chert Patination as a Dating Technique: A Case Study from Fort Hood, Texas. Research Report No. 32, Archaeological Resource Management Series. United States Army, Fort Hood, TX. Glascock, Michael D. 94 1992 Characterization of Archaeological Ceramics at MURR by Neutron Activation Analysis and Multivariate Statistics. In Chemical Characterization of Ceramic Pastes in Archaeology, edited by Hector Neff, pp. 11-26. Prehistory Press, Madison, WI. Glascock, Michael D. and Hector Neff 2003 Neutron Activation Analysis and Provenance Research in Archaeology. Measurement Science and Technology 14:1516-1526. Harbottle, Garman 1976 Activation Analysis in Archaeology. Radiochemistry 3(1):33-72. Ives, David J. 1984 Neutron Activation Analysis Characterization of Selected Prehistoric Chert Quarrying Areas. Unpublished Ph.D. dissertation, University of Missouri, Columbia. Leese, Morven N. and Peter L. Main 1994 The Efficient Computation of Unbiased Mahalanobis Distances and their Interpretation in Archaeometry. Archaeometry 36:307-316. Neff, Hector 1994 RQ-mode Principal Component Analysis of Ceramic Compositional Data. Archaeometry 36:115-130. 2000 Neutron Activation Analysis for Provenance Determination in Archaeology. In Modern Analytical Methods in Art and Archaeology, edited by Enrico Ciliberto and Giuseppe Spoto, pp. 81-134. John Wiley and Sons, New York. Sayre, Edward V. 1975 Brookhaven Procedures for Statistical Analyses of Multivariate Archaeometric Data. Brookhaven National Laboratory Report BNL-23128. Steponaitis, Vincas, M. James Blackman and Hector Neff 1996 Large-scale Compositional Patterns in the Chemical Composition of Mississippian Pottery. American Antiquity 61(3):555-572. Weigand, Phil C., Garman Harbottle and Edward V. Sayre 1977 Turquoise Sources and Source Analysis: Mesoamerica and the Southwestern U.S.A. In Exchange Systems in Prehistory, edited by Timothy K. Earle and J. E. Ericson, pp. 15-34. Academic Press, New York. 95 Table 2. Principal components analysis of novaculite samples obtained from the Texas, Arkansas, and Oklahoma. Note that greater than 90% of the total variance in the dataset is explained by these five principal components. Eigenvalue % Variance % Cum. Variance Ba La Nd Sm U Yb Ce Co Cr Cs Eu Fe Hf Rb Sb Sc Sr Ta Tb Th Zn Al Mn Na 1 4.4733 72.1757 72.1757 2 0.539 8.6971 80.8728 3 0.3236 5.2216 86.0944 4 0.2403 3.8766 89.971 5 0.1529 2.4676 92.4387 -0.0225 -0.1117 -0.1087 -0.0616 0.245 -0.302 -0.147 -0.2379 -0.072 -0.2783 -0.3463 -0.1888 -0.2321 -0.1857 -0.1546 -0.2579 -0.1466 -0.1633 -0.313 -0.2057 -0.2403 -0.0165 -0.2785 0.0299 0.1324 0.2736 0.2467 0.231 0.156 0.0854 0.2008 -0.3956 0.0108 -0.0666 0.1595 0.0147 0.1451 0.0837 0.1725 -0.0017 0.1688 0.0807 0.1207 0.0618 -0.1023 0.0409 -0.641 -0.0224 0.0649 0.1602 0.1712 0.1659 0.0962 -0.0744 0.17 -0.3509 -0.0645 -0.1074 0.0181 -0.2108 0.0436 -0.0316 -0.4765 0.0535 0.4532 -0.0362 -0.0166 -0.0514 -0.0026 0.0651 0.4879 0.0359 -0.2967 -0.1166 -0.1536 -0.2097 -0.5451 0.1728 -0.1375 -0.4498 0.0334 -0.0748 0.0788 -0.3188 -0.0469 -0.1542 -0.1031 0.1788 -0.0353 0.0037 0.1695 0.0589 0.1239 -0.0815 -0.1503 -0.1462 -0.2838 -0.0366 0.005 0.0179 0.0768 0.0252 -0.0906 -0.071 0.0854 -0.3781 0.0361 0.5423 -0.1835 -0.3746 0.2966 0.1237 0.3586 -0.1575 0.0706 -0.0571 0.0152 -0.0488 0.1005 -0.0274 96 Figure 1. Locations of novaculite samples analyzed in this study. A: Big Hudson Creek; B: Caddo Gap; C: 3GA48 D: 3HS69/603 Base data obtained from GeoStor (http://www.geostor.arkansas.gov/) and the University of Oklahoma’s Center for Spatial Analysis (http://www.csa.ou.edu/). 97 3GA48 3HS69/603 3GA48 3HS69/603 3GA48 3HS69/603 3GA48 3HS69/603 Figure 6. Bivariate plot of logged uranium and thorium concentrations showing samples analyzed as part of this study compared with previously analyzed novaculite and chert samples from the region. Ellipses represent 90% confidence interval of group membership. Note that samples from this study, excepting sample MBT015, are generally lower in uranium and higher in REEs such as thorium. 102 Figure 7. Bivariate plot of logged uranium and europium concentrations showing the newly analyzed novaculite samples against all previously analyzed cryptocrystalline silicates (i.e., chert, novaculite, agate, etc.) from Texas and Oklahoma. Data compiled from a number of sources. Ellipse represents 90% confidence interval of group membership. Note that the sample falling within ellipse for the Arkansas and Oklahoma novaculites is a previously analyzed artifact and not a source sample. 103 3GA48 3HS69/603 Appendix A: Descriptive Statistics for Newly Analyzed Novaculite Samples from Arkansas and Oklahoma 105 3GA48 Descriptive Statistics for Novaculite Samples from % St. Dev. e, AR Element Mean St. Dev. No. Obs. Min Max As 0.860 0.458 53.195 3 0.382 1.294 Ba 21.682 11.848 54.647 5 12.451 42.371 La 1.264 0.740 58.507 5 0.486 2.392 Lu 0.012 0.006 51.022 5 0.007 0.022 Nd 1.389 1.199 86.335 5 0.525 3.476 Sm 0.277 0.209 75.519 5 0.131 0.643 U 0.154 0.037 23.902 5 0.097 0.197 Yb 0.086 0.052 60.486 5 0.041 0.175 Ce 1.633 0.871 53.349 5 0.816 2.977 Co 0.011 0.005 41.388 5 0.007 0.019 Cr 0.504 0.097 19.234 5 0.441 0.674 Cs 0.046 0.007 14.756 5 0.038 0.053 Eu 0.059 0.049 83.020 5 0.023 0.144 Fe 402.424 294.932 73.289 5 48.019 835.793 Hf 0.163 0.109 66.761 5 0.064 0.336 Ni 2.070 1 2.070 2.070 Rb 0.703 5 0.578 0.931 . 0.139 . 19.721 Sb 0.066 0.033 49.462 5 0.031 0.106 Sc 0.176 0.064 36.245 5 0.123 0.283 Sr 31.197 49.598 158.982 5 2.223 118.227 Ta 0.009 0.004 40.984 5 0.004 0.014 Tb 0.038 0.033 87.203 5 0.015 0.097 Th 0.139 0.064 45.892 5 0.069 0.240 Zn 1.543 0.959 62.172 5 0.449 2.628 Al 1649.813 663.325 40.206 5 1097.718 2803.493 Ca 606.963 1 606.963 606.963 Dy 0.344 0.316 91.737 5 0.088 0.862 196.466 65.192 33.182 5 126.705 271.684 Mn 1.291 0.532 41.223 5 0.757 2.062 Na 84.449 6.757 8.001 5 78.369 93.406 Ti 32.586 8.960 27.496 3 22.956 40.676 V 2.007 2.772 138.111 5 0.561 6.959 K . . ANIDS of specimens included: MBT001 MBT002 MBT003 MBT004 MBT005 106 3HS69/603 Descriptive Statistics for Novaculite Samples from AR Element Mean St. Dev. % St. Dev. No. Obs. Min Max As 0.288 0.328 113.852 6 0.026 0.799 Ba 18.172 12.243 67.374 8 3.869 35.387 La 0.572 0.503 88.046 8 0.257 1.772 Lu 0.008 0.006 78.621 8 0.003 0.020 Nd 0.725 0.589 81.302 8 0.197 2.123 Sm 0.161 0.109 67.595 8 0.042 0.414 U 0.109 0.038 35.360 8 0.046 0.177 Yb 0.057 0.045 79.446 8 0.020 0.132 Ce 1.253 1.045 83.412 8 0.528 3.739 Co 0.122 0.200 164.032 8 0.007 0.568 Cr 0.573 0.301 52.544 8 0.353 1.277 Cs 0.072 0.036 50.029 8 0.035 0.131 Eu 0.033 0.022 65.192 8 0.007 0.082 Fe 202.643 187.655 92.604 8 27.505 643.481 Hf 0.096 0.092 95.973 8 0.045 0.321 Ni 1.299 0.483 37.191 2 0.958 1.641 Rb 0.892 0.622 69.700 8 0.119 1.825 Sb 0.024 0.025 104.708 8 0.003 0.081 Sc 0.168 0.123 73.017 8 0.051 0.412 Sr 4.955 4.531 91.448 8 0.733 12.123 Ta 0.006 0.004 60.417 8 0.003 0.014 Tb 0.022 0.014 65.370 8 0.004 0.052 Th 0.108 0.078 72.497 8 0.054 0.290 Zn 2.470 3.067 124.170 8 0.623 8.276 Al 1389.656 280.435 20.180 8 984.364 1758.813 Ca 1330.401 1156.950 86.962 5 45.018 2569.974 Dy 0.117 0.086 73.532 8 0.036 0.314 295.692 180.630 61.087 7 102.762 606.325 Mn 48.234 62.073 128.691 8 0.327 176.445 Na 101.476 21.318 21.008 8 73.377 133.990 K Ti V . . 1.440 0.910 . 1 63.226 5 ANIDS of specimens included: MBT006 MBT007 MBT008 MBT009 MBT010 MBT011 MBT012 MBT013 107 . 0.370 . 2.794 Descriptive Statistics for Novaculite Samples from Caddo Gap, AR Element Mean St. Dev. % St. Dev. No. Obs. Min Max As 1.455 2.245 154.247 4 0.102 4.809 Ba 16.701 10.734 64.270 5 7.344 34.809 La 1.196 1.372 114.679 5 0.447 3.618 Lu 0.016 0.008 50.346 5 0.006 0.026 Nd 1.427 1.403 98.285 5 0.452 3.891 Sm 0.326 0.289 88.618 5 0.085 0.824 U 0.282 0.281 99.598 5 0.097 0.776 Yb 0.118 0.066 55.717 5 0.039 0.199 Ce 1.483 1.419 95.679 5 0.523 3.985 Co 0.181 0.156 86.018 5 0.012 0.345 Cr 0.896 0.221 24.718 5 0.635 1.117 Cs 0.106 0.089 84.084 5 0.035 0.259 Eu 0.071 0.066 93.709 5 0.015 0.185 Fe 577.141 753.057 130.481 5 47.349 1908.275 Hf 0.123 0.111 89.992 5 0.029 0.304 Ni 1.482 0.524 35.375 3 1.158 2.086 Rb 0.967 0.513 53.058 5 0.462 1.794 Sb 0.074 0.052 70.727 5 0.012 0.147 Sc 0.261 0.205 78.366 5 0.081 0.598 Sr 5.877 6.429 109.391 4 1.119 15.276 Ta 0.008 0.005 61.340 5 0.004 0.017 Tb 0.051 0.044 85.982 5 0.009 0.125 Th 0.165 0.118 71.511 5 0.103 0.374 Zn 0.947 0.734 77.543 5 0.240 2.070 Al 1503.638 399.658 26.579 5 1084.843 2049.830 Ca 122.660 1 122.660 122.660 Dy 0.300 0.231 77.023 5 0.041 0.624 229.904 136.956 59.571 5 63.291 409.294 Mn 1.502 1.340 89.254 5 0.642 3.877 Na 118.177 53.477 45.252 5 83.487 212.196 Ti 55.142 29.654 53.777 3 26.846 85.989 V 1.680 0.661 39.335 5 1.063 2.664 K . . ANIDS of specimens included: MBT014 MBT015 MBT016 MBT017 MBT018 108 Descriptive Statistics for Novaculite Samples from Big Hudson Creek, OK. Element Mean St. Dev. % St. Dev. No. Obs. Min Max As 0.164 0.124 75.620 2 0.076 0.252 Ba 40.450 2.425 5.994 2 38.735 42.164 La 2.017 1.326 65.733 2 1.080 2.955 Lu 0.010 0.002 17.849 2 0.009 0.012 Nd 2.047 1.265 61.809 2 1.152 2.942 Sm 0.384 0.180 46.956 2 0.257 0.512 U 0.151 0.054 35.777 2 0.113 0.189 Yb 0.083 0.001 1.103 2 0.083 0.084 Ce 5.778 5.150 89.135 2 2.136 9.420 Co 0.019 0.005 28.284 2 0.015 0.022 Cr 1.366 1.082 79.227 2 0.601 2.131 Cs 0.050 0.023 46.097 2 0.033 0.066 Eu 0.078 0.028 36.443 2 0.058 0.098 Fe 271.426 87.009 32.056 2 209.901 332.951 Hf 0.109 0.053 49.043 2 0.071 0.147 Ni 0.570 1 0.570 0.570 Rb 1.819 2 1.016 2.622 . . 1.136 62.442 Sb 0.021 0.004 19.541 2 0.018 0.024 Sc 0.299 0.024 8.163 2 0.282 0.316 Sr 7.574 4.315 56.969 2 4.523 10.625 Ta 0.011 0.008 73.839 2 0.005 0.017 Tb 0.043 0.013 30.107 2 0.034 0.053 Th 0.171 0.045 26.501 2 0.139 0.203 Zn 1.938 0.010 0.529 2 1.931 1.946 Al 1691.944 259.526 15.339 2 1508.432 1875.457 Ca Dy . . . 1 . . 0.211 0.117 55.602 2 0.128 0.293 368.083 199.974 54.329 2 226.680 509.486 Mn 1.598 0.090 5.643 2 1.534 1.661 Na 65.905 4.010 6.084 2 63.070 68.740 Ti 55.608 19.526 35.114 2 41.801 69.415 V 1.215 0.607 49.929 2 0.786 1.645 K ANIDS of specimens included: MBT019 MBT020 109 Appendix B: Chemical Data for Newly Analyzed Novaculite Samples from Arkansas and Oklahoma 110 111 112